There are three main mechanisms of breakage in Autogenous Grinding (AG) and Semi-Autogenous Grinding (SAG) mills:
(ii) Chipping, and
Abrasion is a surface event in which small grains of particles are removed. Chipping can also be classified as surface event in which surface "features" are removed. During both events cracks are not propagated through the main body of the rock and therefore can be classified as surface breakage. However, In crushing cracks propagate through the body of the rock.
In modelling breakage is described by using a so-called appearance/breakage function which historically has made use of size-average breakage parameters (Herbst and Fuerstenau (1968); Austin and Luckie (1971); Leung (1987)), ie rock strength is assumed to be independent of size. In addition these parameters typically do not consider the mode of breakage or as in the case of Leung do so in an arbitrary manner. Given that in AG/SAG mills the feed size distribution is usually -250 mm and the product size P80 can be as low as 50 microns the assumption of sizeinvariant strength is likely to have a significant impact on the model's accuracy.
To rectify the current lack of appropriate definition in the way that rock breakage is described in existing AG/SAG mill models, a research program was undertaken via the AM IRA POL project. This thesis describes the resultant model which Incorporates the effect of particle size for the two main regimes of breakage, namely surface and crushing, and unifies them in a single, mechanistically sound, description of breakage.
To quantify surface breakage at different energy levels, a range of equipment was developed. This equipment allows the fragments generated for each breakage event to be extracted. Using four different mill diameters, the effect of input energy and particle size on mass of fines produced was determined. The results showed that under very low Input energy conditions, which produce chipping and abrasion breakage, the resultant product size and quantity of broken products are affected by the size of the particles. As expected, the coarser particles tend to generate more fragments than the smaller ones. Although the common method of relating the amount of fragments produced during breakage is by the use of mass specific input energy, due to the fact that abrasion and chipping is related to surface conditions, the data were found to be correlated to surface specific input energy instead.
A drop weight tester was used to quantify the effect of particle strength on the particle size distribution obtained under higher energy impact breakage conditions (eg as observed in crushing). A wide range of particle sizes was tested under a wide range of input energies. As in the case of surface breakage, the results showed that coarser particles tend to be inherently weaker than smaller ones. In view of the fact that during crushing cracks propagate through the body of the rock, the effect observed was modelled based on the volumetric specific input energy.
The above descriptions of "surface" (eg abrasion and chipping) and "body" (eg crushing) breakage were integrated into a single description of breakage for use in AG/SAG mill modelling. To do so the size distributions of products from these two types of breakage were combined using a weighting function. To determine the weighting, the Hopkinson Pressure Bar (HPB) was used to further characterise the rocks. The data from the HPB was modelled using the log-normal distribution. The approach provides a probability distribution that allows the determination of the proportion of the material that will break either by body or surface breakage under a given input energy.
The breakage model has been incorporated in a prototype version of a new AG/SAG mill model. Evaluation of the model has shown that it can mimic expected trends during SAG mill operation. Similar simulation studies were carried out using the current (Leung's) appearance function technique and the results showed that the trends obtained are in contradiction to operational experience. This apparent discrepancy further highlights the limitation of the current techniques for describing breakage.